skbuff.h 34 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910911912913914915916917918919920921922923924925926927928929930931932933934935936937938939940941942943944945946947948949950951952953954955956957958959960961962963964965966967968969970971972973974975976977978979980981982983984985986987988989990991992993994995996997998999100010011002100310041005100610071008100910101011101210131014101510161017101810191020102110221023102410251026102710281029103010311032103310341035103610371038103910401041104210431044104510461047104810491050105110521053105410551056105710581059106010611062106310641065106610671068106910701071107210731074107510761077107810791080108110821083108410851086108710881089109010911092109310941095109610971098109911001101110211031104110511061107110811091110111111121113111411151116111711181119112011211122112311241125112611271128112911301131113211331134113511361137113811391140114111421143114411451146114711481149115011511152115311541155115611571158115911601161116211631164116511661167116811691170117111721173117411751176117711781179118011811182118311841185118611871188118911901191119211931194119511961197119811991200120112021203120412051206120712081209121012111212121312141215121612171218121912201221122212231224122512261227122812291230123112321233123412351236123712381239124012411242124312441245124612471248124912501251125212531254125512561257125812591260126112621263126412651266
  1. /*
  2. * Definitions for the 'struct sk_buff' memory handlers.
  3. *
  4. * Authors:
  5. * Alan Cox, <gw4pts@gw4pts.ampr.org>
  6. * Florian La Roche, <rzsfl@rz.uni-sb.de>
  7. *
  8. * This program is free software; you can redistribute it and/or
  9. * modify it under the terms of the GNU General Public License
  10. * as published by the Free Software Foundation; either version
  11. * 2 of the License, or (at your option) any later version.
  12. */
  13. #ifndef _LINUX_SKBUFF_H
  14. #define _LINUX_SKBUFF_H
  15. #include <linux/config.h>
  16. #include <linux/kernel.h>
  17. #include <linux/compiler.h>
  18. #include <linux/time.h>
  19. #include <linux/cache.h>
  20. #include <asm/atomic.h>
  21. #include <asm/types.h>
  22. #include <linux/spinlock.h>
  23. #include <linux/mm.h>
  24. #include <linux/highmem.h>
  25. #include <linux/poll.h>
  26. #include <linux/net.h>
  27. #include <linux/textsearch.h>
  28. #include <net/checksum.h>
  29. #define HAVE_ALLOC_SKB /* For the drivers to know */
  30. #define HAVE_ALIGNABLE_SKB /* Ditto 8) */
  31. #define SLAB_SKB /* Slabified skbuffs */
  32. #define CHECKSUM_NONE 0
  33. #define CHECKSUM_HW 1
  34. #define CHECKSUM_UNNECESSARY 2
  35. #define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
  36. ~(SMP_CACHE_BYTES - 1))
  37. #define SKB_MAX_ORDER(X, ORDER) (((PAGE_SIZE << (ORDER)) - (X) - \
  38. sizeof(struct skb_shared_info)) & \
  39. ~(SMP_CACHE_BYTES - 1))
  40. #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
  41. #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
  42. /* A. Checksumming of received packets by device.
  43. *
  44. * NONE: device failed to checksum this packet.
  45. * skb->csum is undefined.
  46. *
  47. * UNNECESSARY: device parsed packet and wouldbe verified checksum.
  48. * skb->csum is undefined.
  49. * It is bad option, but, unfortunately, many of vendors do this.
  50. * Apparently with secret goal to sell you new device, when you
  51. * will add new protocol to your host. F.e. IPv6. 8)
  52. *
  53. * HW: the most generic way. Device supplied checksum of _all_
  54. * the packet as seen by netif_rx in skb->csum.
  55. * NOTE: Even if device supports only some protocols, but
  56. * is able to produce some skb->csum, it MUST use HW,
  57. * not UNNECESSARY.
  58. *
  59. * B. Checksumming on output.
  60. *
  61. * NONE: skb is checksummed by protocol or csum is not required.
  62. *
  63. * HW: device is required to csum packet as seen by hard_start_xmit
  64. * from skb->h.raw to the end and to record the checksum
  65. * at skb->h.raw+skb->csum.
  66. *
  67. * Device must show its capabilities in dev->features, set
  68. * at device setup time.
  69. * NETIF_F_HW_CSUM - it is clever device, it is able to checksum
  70. * everything.
  71. * NETIF_F_NO_CSUM - loopback or reliable single hop media.
  72. * NETIF_F_IP_CSUM - device is dumb. It is able to csum only
  73. * TCP/UDP over IPv4. Sigh. Vendors like this
  74. * way by an unknown reason. Though, see comment above
  75. * about CHECKSUM_UNNECESSARY. 8)
  76. *
  77. * Any questions? No questions, good. --ANK
  78. */
  79. struct net_device;
  80. #ifdef CONFIG_NETFILTER
  81. struct nf_conntrack {
  82. atomic_t use;
  83. void (*destroy)(struct nf_conntrack *);
  84. };
  85. #ifdef CONFIG_BRIDGE_NETFILTER
  86. struct nf_bridge_info {
  87. atomic_t use;
  88. struct net_device *physindev;
  89. struct net_device *physoutdev;
  90. #if defined(CONFIG_VLAN_8021Q) || defined(CONFIG_VLAN_8021Q_MODULE)
  91. struct net_device *netoutdev;
  92. #endif
  93. unsigned int mask;
  94. unsigned long data[32 / sizeof(unsigned long)];
  95. };
  96. #endif
  97. #endif
  98. struct sk_buff_head {
  99. /* These two members must be first. */
  100. struct sk_buff *next;
  101. struct sk_buff *prev;
  102. __u32 qlen;
  103. spinlock_t lock;
  104. };
  105. struct sk_buff;
  106. /* To allow 64K frame to be packed as single skb without frag_list */
  107. #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 2)
  108. typedef struct skb_frag_struct skb_frag_t;
  109. struct skb_frag_struct {
  110. struct page *page;
  111. __u16 page_offset;
  112. __u16 size;
  113. };
  114. /* This data is invariant across clones and lives at
  115. * the end of the header data, ie. at skb->end.
  116. */
  117. struct skb_shared_info {
  118. atomic_t dataref;
  119. unsigned int nr_frags;
  120. unsigned short tso_size;
  121. unsigned short tso_segs;
  122. struct sk_buff *frag_list;
  123. skb_frag_t frags[MAX_SKB_FRAGS];
  124. };
  125. /* We divide dataref into two halves. The higher 16 bits hold references
  126. * to the payload part of skb->data. The lower 16 bits hold references to
  127. * the entire skb->data. It is up to the users of the skb to agree on
  128. * where the payload starts.
  129. *
  130. * All users must obey the rule that the skb->data reference count must be
  131. * greater than or equal to the payload reference count.
  132. *
  133. * Holding a reference to the payload part means that the user does not
  134. * care about modifications to the header part of skb->data.
  135. */
  136. #define SKB_DATAREF_SHIFT 16
  137. #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
  138. /**
  139. * struct sk_buff - socket buffer
  140. * @next: Next buffer in list
  141. * @prev: Previous buffer in list
  142. * @list: List we are on
  143. * @sk: Socket we are owned by
  144. * @stamp: Time we arrived
  145. * @dev: Device we arrived on/are leaving by
  146. * @input_dev: Device we arrived on
  147. * @real_dev: The real device we are using
  148. * @h: Transport layer header
  149. * @nh: Network layer header
  150. * @mac: Link layer header
  151. * @dst: destination entry
  152. * @sp: the security path, used for xfrm
  153. * @cb: Control buffer. Free for use by every layer. Put private vars here
  154. * @len: Length of actual data
  155. * @data_len: Data length
  156. * @mac_len: Length of link layer header
  157. * @csum: Checksum
  158. * @local_df: allow local fragmentation
  159. * @cloned: Head may be cloned (check refcnt to be sure)
  160. * @nohdr: Payload reference only, must not modify header
  161. * @pkt_type: Packet class
  162. * @ip_summed: Driver fed us an IP checksum
  163. * @priority: Packet queueing priority
  164. * @users: User count - see {datagram,tcp}.c
  165. * @protocol: Packet protocol from driver
  166. * @truesize: Buffer size
  167. * @head: Head of buffer
  168. * @data: Data head pointer
  169. * @tail: Tail pointer
  170. * @end: End pointer
  171. * @destructor: Destruct function
  172. * @nfmark: Can be used for communication between hooks
  173. * @nfcache: Cache info
  174. * @nfct: Associated connection, if any
  175. * @nfctinfo: Relationship of this skb to the connection
  176. * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
  177. * @private: Data which is private to the HIPPI implementation
  178. * @tc_index: Traffic control index
  179. * @tc_verd: traffic control verdict
  180. * @tc_classid: traffic control classid
  181. */
  182. struct sk_buff {
  183. /* These two members must be first. */
  184. struct sk_buff *next;
  185. struct sk_buff *prev;
  186. struct sk_buff_head *list;
  187. struct sock *sk;
  188. struct timeval stamp;
  189. struct net_device *dev;
  190. struct net_device *input_dev;
  191. struct net_device *real_dev;
  192. union {
  193. struct tcphdr *th;
  194. struct udphdr *uh;
  195. struct icmphdr *icmph;
  196. struct igmphdr *igmph;
  197. struct iphdr *ipiph;
  198. struct ipv6hdr *ipv6h;
  199. unsigned char *raw;
  200. } h;
  201. union {
  202. struct iphdr *iph;
  203. struct ipv6hdr *ipv6h;
  204. struct arphdr *arph;
  205. unsigned char *raw;
  206. } nh;
  207. union {
  208. unsigned char *raw;
  209. } mac;
  210. struct dst_entry *dst;
  211. struct sec_path *sp;
  212. /*
  213. * This is the control buffer. It is free to use for every
  214. * layer. Please put your private variables there. If you
  215. * want to keep them across layers you have to do a skb_clone()
  216. * first. This is owned by whoever has the skb queued ATM.
  217. */
  218. char cb[40];
  219. unsigned int len,
  220. data_len,
  221. mac_len,
  222. csum;
  223. __u32 priority;
  224. __u8 local_df:1,
  225. cloned:1,
  226. ip_summed:2,
  227. nohdr:1;
  228. /* 3 bits spare */
  229. __u8 pkt_type;
  230. __u16 protocol;
  231. void (*destructor)(struct sk_buff *skb);
  232. #ifdef CONFIG_NETFILTER
  233. unsigned long nfmark;
  234. __u32 nfcache;
  235. __u32 nfctinfo;
  236. struct nf_conntrack *nfct;
  237. #ifdef CONFIG_BRIDGE_NETFILTER
  238. struct nf_bridge_info *nf_bridge;
  239. #endif
  240. #endif /* CONFIG_NETFILTER */
  241. #if defined(CONFIG_HIPPI)
  242. union {
  243. __u32 ifield;
  244. } private;
  245. #endif
  246. #ifdef CONFIG_NET_SCHED
  247. __u32 tc_index; /* traffic control index */
  248. #ifdef CONFIG_NET_CLS_ACT
  249. __u32 tc_verd; /* traffic control verdict */
  250. __u32 tc_classid; /* traffic control classid */
  251. #endif
  252. #endif
  253. /* These elements must be at the end, see alloc_skb() for details. */
  254. unsigned int truesize;
  255. atomic_t users;
  256. unsigned char *head,
  257. *data,
  258. *tail,
  259. *end;
  260. };
  261. #ifdef __KERNEL__
  262. /*
  263. * Handling routines are only of interest to the kernel
  264. */
  265. #include <linux/slab.h>
  266. #include <asm/system.h>
  267. extern void __kfree_skb(struct sk_buff *skb);
  268. extern struct sk_buff *alloc_skb(unsigned int size,
  269. unsigned int __nocast priority);
  270. extern struct sk_buff *alloc_skb_from_cache(kmem_cache_t *cp,
  271. unsigned int size,
  272. unsigned int __nocast priority);
  273. extern void kfree_skbmem(struct sk_buff *skb);
  274. extern struct sk_buff *skb_clone(struct sk_buff *skb,
  275. unsigned int __nocast priority);
  276. extern struct sk_buff *skb_copy(const struct sk_buff *skb,
  277. unsigned int __nocast priority);
  278. extern struct sk_buff *pskb_copy(struct sk_buff *skb,
  279. unsigned int __nocast gfp_mask);
  280. extern int pskb_expand_head(struct sk_buff *skb,
  281. int nhead, int ntail,
  282. unsigned int __nocast gfp_mask);
  283. extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
  284. unsigned int headroom);
  285. extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
  286. int newheadroom, int newtailroom,
  287. unsigned int __nocast priority);
  288. extern struct sk_buff * skb_pad(struct sk_buff *skb, int pad);
  289. #define dev_kfree_skb(a) kfree_skb(a)
  290. extern void skb_over_panic(struct sk_buff *skb, int len,
  291. void *here);
  292. extern void skb_under_panic(struct sk_buff *skb, int len,
  293. void *here);
  294. struct skb_seq_state
  295. {
  296. __u32 lower_offset;
  297. __u32 upper_offset;
  298. __u32 frag_idx;
  299. __u32 stepped_offset;
  300. struct sk_buff *root_skb;
  301. struct sk_buff *cur_skb;
  302. __u8 *frag_data;
  303. };
  304. extern void skb_prepare_seq_read(struct sk_buff *skb,
  305. unsigned int from, unsigned int to,
  306. struct skb_seq_state *st);
  307. extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
  308. struct skb_seq_state *st);
  309. extern void skb_abort_seq_read(struct skb_seq_state *st);
  310. extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
  311. unsigned int to, struct ts_config *config,
  312. struct ts_state *state);
  313. /* Internal */
  314. #define skb_shinfo(SKB) ((struct skb_shared_info *)((SKB)->end))
  315. /**
  316. * skb_queue_empty - check if a queue is empty
  317. * @list: queue head
  318. *
  319. * Returns true if the queue is empty, false otherwise.
  320. */
  321. static inline int skb_queue_empty(const struct sk_buff_head *list)
  322. {
  323. return list->next == (struct sk_buff *)list;
  324. }
  325. /**
  326. * skb_get - reference buffer
  327. * @skb: buffer to reference
  328. *
  329. * Makes another reference to a socket buffer and returns a pointer
  330. * to the buffer.
  331. */
  332. static inline struct sk_buff *skb_get(struct sk_buff *skb)
  333. {
  334. atomic_inc(&skb->users);
  335. return skb;
  336. }
  337. /*
  338. * If users == 1, we are the only owner and are can avoid redundant
  339. * atomic change.
  340. */
  341. /**
  342. * kfree_skb - free an sk_buff
  343. * @skb: buffer to free
  344. *
  345. * Drop a reference to the buffer and free it if the usage count has
  346. * hit zero.
  347. */
  348. static inline void kfree_skb(struct sk_buff *skb)
  349. {
  350. if (likely(atomic_read(&skb->users) == 1))
  351. smp_rmb();
  352. else if (likely(!atomic_dec_and_test(&skb->users)))
  353. return;
  354. __kfree_skb(skb);
  355. }
  356. /**
  357. * skb_cloned - is the buffer a clone
  358. * @skb: buffer to check
  359. *
  360. * Returns true if the buffer was generated with skb_clone() and is
  361. * one of multiple shared copies of the buffer. Cloned buffers are
  362. * shared data so must not be written to under normal circumstances.
  363. */
  364. static inline int skb_cloned(const struct sk_buff *skb)
  365. {
  366. return skb->cloned &&
  367. (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
  368. }
  369. /**
  370. * skb_header_cloned - is the header a clone
  371. * @skb: buffer to check
  372. *
  373. * Returns true if modifying the header part of the buffer requires
  374. * the data to be copied.
  375. */
  376. static inline int skb_header_cloned(const struct sk_buff *skb)
  377. {
  378. int dataref;
  379. if (!skb->cloned)
  380. return 0;
  381. dataref = atomic_read(&skb_shinfo(skb)->dataref);
  382. dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
  383. return dataref != 1;
  384. }
  385. /**
  386. * skb_header_release - release reference to header
  387. * @skb: buffer to operate on
  388. *
  389. * Drop a reference to the header part of the buffer. This is done
  390. * by acquiring a payload reference. You must not read from the header
  391. * part of skb->data after this.
  392. */
  393. static inline void skb_header_release(struct sk_buff *skb)
  394. {
  395. BUG_ON(skb->nohdr);
  396. skb->nohdr = 1;
  397. atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
  398. }
  399. /**
  400. * skb_shared - is the buffer shared
  401. * @skb: buffer to check
  402. *
  403. * Returns true if more than one person has a reference to this
  404. * buffer.
  405. */
  406. static inline int skb_shared(const struct sk_buff *skb)
  407. {
  408. return atomic_read(&skb->users) != 1;
  409. }
  410. /**
  411. * skb_share_check - check if buffer is shared and if so clone it
  412. * @skb: buffer to check
  413. * @pri: priority for memory allocation
  414. *
  415. * If the buffer is shared the buffer is cloned and the old copy
  416. * drops a reference. A new clone with a single reference is returned.
  417. * If the buffer is not shared the original buffer is returned. When
  418. * being called from interrupt status or with spinlocks held pri must
  419. * be GFP_ATOMIC.
  420. *
  421. * NULL is returned on a memory allocation failure.
  422. */
  423. static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
  424. unsigned int __nocast pri)
  425. {
  426. might_sleep_if(pri & __GFP_WAIT);
  427. if (skb_shared(skb)) {
  428. struct sk_buff *nskb = skb_clone(skb, pri);
  429. kfree_skb(skb);
  430. skb = nskb;
  431. }
  432. return skb;
  433. }
  434. /*
  435. * Copy shared buffers into a new sk_buff. We effectively do COW on
  436. * packets to handle cases where we have a local reader and forward
  437. * and a couple of other messy ones. The normal one is tcpdumping
  438. * a packet thats being forwarded.
  439. */
  440. /**
  441. * skb_unshare - make a copy of a shared buffer
  442. * @skb: buffer to check
  443. * @pri: priority for memory allocation
  444. *
  445. * If the socket buffer is a clone then this function creates a new
  446. * copy of the data, drops a reference count on the old copy and returns
  447. * the new copy with the reference count at 1. If the buffer is not a clone
  448. * the original buffer is returned. When called with a spinlock held or
  449. * from interrupt state @pri must be %GFP_ATOMIC
  450. *
  451. * %NULL is returned on a memory allocation failure.
  452. */
  453. static inline struct sk_buff *skb_unshare(struct sk_buff *skb, int pri)
  454. {
  455. might_sleep_if(pri & __GFP_WAIT);
  456. if (skb_cloned(skb)) {
  457. struct sk_buff *nskb = skb_copy(skb, pri);
  458. kfree_skb(skb); /* Free our shared copy */
  459. skb = nskb;
  460. }
  461. return skb;
  462. }
  463. /**
  464. * skb_peek
  465. * @list_: list to peek at
  466. *
  467. * Peek an &sk_buff. Unlike most other operations you _MUST_
  468. * be careful with this one. A peek leaves the buffer on the
  469. * list and someone else may run off with it. You must hold
  470. * the appropriate locks or have a private queue to do this.
  471. *
  472. * Returns %NULL for an empty list or a pointer to the head element.
  473. * The reference count is not incremented and the reference is therefore
  474. * volatile. Use with caution.
  475. */
  476. static inline struct sk_buff *skb_peek(struct sk_buff_head *list_)
  477. {
  478. struct sk_buff *list = ((struct sk_buff *)list_)->next;
  479. if (list == (struct sk_buff *)list_)
  480. list = NULL;
  481. return list;
  482. }
  483. /**
  484. * skb_peek_tail
  485. * @list_: list to peek at
  486. *
  487. * Peek an &sk_buff. Unlike most other operations you _MUST_
  488. * be careful with this one. A peek leaves the buffer on the
  489. * list and someone else may run off with it. You must hold
  490. * the appropriate locks or have a private queue to do this.
  491. *
  492. * Returns %NULL for an empty list or a pointer to the tail element.
  493. * The reference count is not incremented and the reference is therefore
  494. * volatile. Use with caution.
  495. */
  496. static inline struct sk_buff *skb_peek_tail(struct sk_buff_head *list_)
  497. {
  498. struct sk_buff *list = ((struct sk_buff *)list_)->prev;
  499. if (list == (struct sk_buff *)list_)
  500. list = NULL;
  501. return list;
  502. }
  503. /**
  504. * skb_queue_len - get queue length
  505. * @list_: list to measure
  506. *
  507. * Return the length of an &sk_buff queue.
  508. */
  509. static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
  510. {
  511. return list_->qlen;
  512. }
  513. static inline void skb_queue_head_init(struct sk_buff_head *list)
  514. {
  515. spin_lock_init(&list->lock);
  516. list->prev = list->next = (struct sk_buff *)list;
  517. list->qlen = 0;
  518. }
  519. /*
  520. * Insert an sk_buff at the start of a list.
  521. *
  522. * The "__skb_xxxx()" functions are the non-atomic ones that
  523. * can only be called with interrupts disabled.
  524. */
  525. /**
  526. * __skb_queue_head - queue a buffer at the list head
  527. * @list: list to use
  528. * @newsk: buffer to queue
  529. *
  530. * Queue a buffer at the start of a list. This function takes no locks
  531. * and you must therefore hold required locks before calling it.
  532. *
  533. * A buffer cannot be placed on two lists at the same time.
  534. */
  535. extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
  536. static inline void __skb_queue_head(struct sk_buff_head *list,
  537. struct sk_buff *newsk)
  538. {
  539. struct sk_buff *prev, *next;
  540. newsk->list = list;
  541. list->qlen++;
  542. prev = (struct sk_buff *)list;
  543. next = prev->next;
  544. newsk->next = next;
  545. newsk->prev = prev;
  546. next->prev = prev->next = newsk;
  547. }
  548. /**
  549. * __skb_queue_tail - queue a buffer at the list tail
  550. * @list: list to use
  551. * @newsk: buffer to queue
  552. *
  553. * Queue a buffer at the end of a list. This function takes no locks
  554. * and you must therefore hold required locks before calling it.
  555. *
  556. * A buffer cannot be placed on two lists at the same time.
  557. */
  558. extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
  559. static inline void __skb_queue_tail(struct sk_buff_head *list,
  560. struct sk_buff *newsk)
  561. {
  562. struct sk_buff *prev, *next;
  563. newsk->list = list;
  564. list->qlen++;
  565. next = (struct sk_buff *)list;
  566. prev = next->prev;
  567. newsk->next = next;
  568. newsk->prev = prev;
  569. next->prev = prev->next = newsk;
  570. }
  571. /**
  572. * __skb_dequeue - remove from the head of the queue
  573. * @list: list to dequeue from
  574. *
  575. * Remove the head of the list. This function does not take any locks
  576. * so must be used with appropriate locks held only. The head item is
  577. * returned or %NULL if the list is empty.
  578. */
  579. extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
  580. static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
  581. {
  582. struct sk_buff *next, *prev, *result;
  583. prev = (struct sk_buff *) list;
  584. next = prev->next;
  585. result = NULL;
  586. if (next != prev) {
  587. result = next;
  588. next = next->next;
  589. list->qlen--;
  590. next->prev = prev;
  591. prev->next = next;
  592. result->next = result->prev = NULL;
  593. result->list = NULL;
  594. }
  595. return result;
  596. }
  597. /*
  598. * Insert a packet on a list.
  599. */
  600. extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk);
  601. static inline void __skb_insert(struct sk_buff *newsk,
  602. struct sk_buff *prev, struct sk_buff *next,
  603. struct sk_buff_head *list)
  604. {
  605. newsk->next = next;
  606. newsk->prev = prev;
  607. next->prev = prev->next = newsk;
  608. newsk->list = list;
  609. list->qlen++;
  610. }
  611. /*
  612. * Place a packet after a given packet in a list.
  613. */
  614. extern void skb_append(struct sk_buff *old, struct sk_buff *newsk);
  615. static inline void __skb_append(struct sk_buff *old, struct sk_buff *newsk)
  616. {
  617. __skb_insert(newsk, old, old->next, old->list);
  618. }
  619. /*
  620. * remove sk_buff from list. _Must_ be called atomically, and with
  621. * the list known..
  622. */
  623. extern void skb_unlink(struct sk_buff *skb);
  624. static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
  625. {
  626. struct sk_buff *next, *prev;
  627. list->qlen--;
  628. next = skb->next;
  629. prev = skb->prev;
  630. skb->next = skb->prev = NULL;
  631. skb->list = NULL;
  632. next->prev = prev;
  633. prev->next = next;
  634. }
  635. /* XXX: more streamlined implementation */
  636. /**
  637. * __skb_dequeue_tail - remove from the tail of the queue
  638. * @list: list to dequeue from
  639. *
  640. * Remove the tail of the list. This function does not take any locks
  641. * so must be used with appropriate locks held only. The tail item is
  642. * returned or %NULL if the list is empty.
  643. */
  644. extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
  645. static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
  646. {
  647. struct sk_buff *skb = skb_peek_tail(list);
  648. if (skb)
  649. __skb_unlink(skb, list);
  650. return skb;
  651. }
  652. static inline int skb_is_nonlinear(const struct sk_buff *skb)
  653. {
  654. return skb->data_len;
  655. }
  656. static inline unsigned int skb_headlen(const struct sk_buff *skb)
  657. {
  658. return skb->len - skb->data_len;
  659. }
  660. static inline int skb_pagelen(const struct sk_buff *skb)
  661. {
  662. int i, len = 0;
  663. for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
  664. len += skb_shinfo(skb)->frags[i].size;
  665. return len + skb_headlen(skb);
  666. }
  667. static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
  668. struct page *page, int off, int size)
  669. {
  670. skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
  671. frag->page = page;
  672. frag->page_offset = off;
  673. frag->size = size;
  674. skb_shinfo(skb)->nr_frags = i + 1;
  675. }
  676. #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
  677. #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_shinfo(skb)->frag_list)
  678. #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
  679. /*
  680. * Add data to an sk_buff
  681. */
  682. static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
  683. {
  684. unsigned char *tmp = skb->tail;
  685. SKB_LINEAR_ASSERT(skb);
  686. skb->tail += len;
  687. skb->len += len;
  688. return tmp;
  689. }
  690. /**
  691. * skb_put - add data to a buffer
  692. * @skb: buffer to use
  693. * @len: amount of data to add
  694. *
  695. * This function extends the used data area of the buffer. If this would
  696. * exceed the total buffer size the kernel will panic. A pointer to the
  697. * first byte of the extra data is returned.
  698. */
  699. static inline unsigned char *skb_put(struct sk_buff *skb, unsigned int len)
  700. {
  701. unsigned char *tmp = skb->tail;
  702. SKB_LINEAR_ASSERT(skb);
  703. skb->tail += len;
  704. skb->len += len;
  705. if (unlikely(skb->tail>skb->end))
  706. skb_over_panic(skb, len, current_text_addr());
  707. return tmp;
  708. }
  709. static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
  710. {
  711. skb->data -= len;
  712. skb->len += len;
  713. return skb->data;
  714. }
  715. /**
  716. * skb_push - add data to the start of a buffer
  717. * @skb: buffer to use
  718. * @len: amount of data to add
  719. *
  720. * This function extends the used data area of the buffer at the buffer
  721. * start. If this would exceed the total buffer headroom the kernel will
  722. * panic. A pointer to the first byte of the extra data is returned.
  723. */
  724. static inline unsigned char *skb_push(struct sk_buff *skb, unsigned int len)
  725. {
  726. skb->data -= len;
  727. skb->len += len;
  728. if (unlikely(skb->data<skb->head))
  729. skb_under_panic(skb, len, current_text_addr());
  730. return skb->data;
  731. }
  732. static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
  733. {
  734. skb->len -= len;
  735. BUG_ON(skb->len < skb->data_len);
  736. return skb->data += len;
  737. }
  738. /**
  739. * skb_pull - remove data from the start of a buffer
  740. * @skb: buffer to use
  741. * @len: amount of data to remove
  742. *
  743. * This function removes data from the start of a buffer, returning
  744. * the memory to the headroom. A pointer to the next data in the buffer
  745. * is returned. Once the data has been pulled future pushes will overwrite
  746. * the old data.
  747. */
  748. static inline unsigned char *skb_pull(struct sk_buff *skb, unsigned int len)
  749. {
  750. return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
  751. }
  752. extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
  753. static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
  754. {
  755. if (len > skb_headlen(skb) &&
  756. !__pskb_pull_tail(skb, len-skb_headlen(skb)))
  757. return NULL;
  758. skb->len -= len;
  759. return skb->data += len;
  760. }
  761. static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
  762. {
  763. return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
  764. }
  765. static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
  766. {
  767. if (likely(len <= skb_headlen(skb)))
  768. return 1;
  769. if (unlikely(len > skb->len))
  770. return 0;
  771. return __pskb_pull_tail(skb, len-skb_headlen(skb)) != NULL;
  772. }
  773. /**
  774. * skb_headroom - bytes at buffer head
  775. * @skb: buffer to check
  776. *
  777. * Return the number of bytes of free space at the head of an &sk_buff.
  778. */
  779. static inline int skb_headroom(const struct sk_buff *skb)
  780. {
  781. return skb->data - skb->head;
  782. }
  783. /**
  784. * skb_tailroom - bytes at buffer end
  785. * @skb: buffer to check
  786. *
  787. * Return the number of bytes of free space at the tail of an sk_buff
  788. */
  789. static inline int skb_tailroom(const struct sk_buff *skb)
  790. {
  791. return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
  792. }
  793. /**
  794. * skb_reserve - adjust headroom
  795. * @skb: buffer to alter
  796. * @len: bytes to move
  797. *
  798. * Increase the headroom of an empty &sk_buff by reducing the tail
  799. * room. This is only allowed for an empty buffer.
  800. */
  801. static inline void skb_reserve(struct sk_buff *skb, unsigned int len)
  802. {
  803. skb->data += len;
  804. skb->tail += len;
  805. }
  806. /*
  807. * CPUs often take a performance hit when accessing unaligned memory
  808. * locations. The actual performance hit varies, it can be small if the
  809. * hardware handles it or large if we have to take an exception and fix it
  810. * in software.
  811. *
  812. * Since an ethernet header is 14 bytes network drivers often end up with
  813. * the IP header at an unaligned offset. The IP header can be aligned by
  814. * shifting the start of the packet by 2 bytes. Drivers should do this
  815. * with:
  816. *
  817. * skb_reserve(NET_IP_ALIGN);
  818. *
  819. * The downside to this alignment of the IP header is that the DMA is now
  820. * unaligned. On some architectures the cost of an unaligned DMA is high
  821. * and this cost outweighs the gains made by aligning the IP header.
  822. *
  823. * Since this trade off varies between architectures, we allow NET_IP_ALIGN
  824. * to be overridden.
  825. */
  826. #ifndef NET_IP_ALIGN
  827. #define NET_IP_ALIGN 2
  828. #endif
  829. extern int ___pskb_trim(struct sk_buff *skb, unsigned int len, int realloc);
  830. static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
  831. {
  832. if (!skb->data_len) {
  833. skb->len = len;
  834. skb->tail = skb->data + len;
  835. } else
  836. ___pskb_trim(skb, len, 0);
  837. }
  838. /**
  839. * skb_trim - remove end from a buffer
  840. * @skb: buffer to alter
  841. * @len: new length
  842. *
  843. * Cut the length of a buffer down by removing data from the tail. If
  844. * the buffer is already under the length specified it is not modified.
  845. */
  846. static inline void skb_trim(struct sk_buff *skb, unsigned int len)
  847. {
  848. if (skb->len > len)
  849. __skb_trim(skb, len);
  850. }
  851. static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
  852. {
  853. if (!skb->data_len) {
  854. skb->len = len;
  855. skb->tail = skb->data+len;
  856. return 0;
  857. }
  858. return ___pskb_trim(skb, len, 1);
  859. }
  860. static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
  861. {
  862. return (len < skb->len) ? __pskb_trim(skb, len) : 0;
  863. }
  864. /**
  865. * skb_orphan - orphan a buffer
  866. * @skb: buffer to orphan
  867. *
  868. * If a buffer currently has an owner then we call the owner's
  869. * destructor function and make the @skb unowned. The buffer continues
  870. * to exist but is no longer charged to its former owner.
  871. */
  872. static inline void skb_orphan(struct sk_buff *skb)
  873. {
  874. if (skb->destructor)
  875. skb->destructor(skb);
  876. skb->destructor = NULL;
  877. skb->sk = NULL;
  878. }
  879. /**
  880. * __skb_queue_purge - empty a list
  881. * @list: list to empty
  882. *
  883. * Delete all buffers on an &sk_buff list. Each buffer is removed from
  884. * the list and one reference dropped. This function does not take the
  885. * list lock and the caller must hold the relevant locks to use it.
  886. */
  887. extern void skb_queue_purge(struct sk_buff_head *list);
  888. static inline void __skb_queue_purge(struct sk_buff_head *list)
  889. {
  890. struct sk_buff *skb;
  891. while ((skb = __skb_dequeue(list)) != NULL)
  892. kfree_skb(skb);
  893. }
  894. #ifndef CONFIG_HAVE_ARCH_DEV_ALLOC_SKB
  895. /**
  896. * __dev_alloc_skb - allocate an skbuff for sending
  897. * @length: length to allocate
  898. * @gfp_mask: get_free_pages mask, passed to alloc_skb
  899. *
  900. * Allocate a new &sk_buff and assign it a usage count of one. The
  901. * buffer has unspecified headroom built in. Users should allocate
  902. * the headroom they think they need without accounting for the
  903. * built in space. The built in space is used for optimisations.
  904. *
  905. * %NULL is returned in there is no free memory.
  906. */
  907. static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
  908. unsigned int __nocast gfp_mask)
  909. {
  910. struct sk_buff *skb = alloc_skb(length + 16, gfp_mask);
  911. if (likely(skb))
  912. skb_reserve(skb, 16);
  913. return skb;
  914. }
  915. #else
  916. extern struct sk_buff *__dev_alloc_skb(unsigned int length, int gfp_mask);
  917. #endif
  918. /**
  919. * dev_alloc_skb - allocate an skbuff for sending
  920. * @length: length to allocate
  921. *
  922. * Allocate a new &sk_buff and assign it a usage count of one. The
  923. * buffer has unspecified headroom built in. Users should allocate
  924. * the headroom they think they need without accounting for the
  925. * built in space. The built in space is used for optimisations.
  926. *
  927. * %NULL is returned in there is no free memory. Although this function
  928. * allocates memory it can be called from an interrupt.
  929. */
  930. static inline struct sk_buff *dev_alloc_skb(unsigned int length)
  931. {
  932. return __dev_alloc_skb(length, GFP_ATOMIC);
  933. }
  934. /**
  935. * skb_cow - copy header of skb when it is required
  936. * @skb: buffer to cow
  937. * @headroom: needed headroom
  938. *
  939. * If the skb passed lacks sufficient headroom or its data part
  940. * is shared, data is reallocated. If reallocation fails, an error
  941. * is returned and original skb is not changed.
  942. *
  943. * The result is skb with writable area skb->head...skb->tail
  944. * and at least @headroom of space at head.
  945. */
  946. static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
  947. {
  948. int delta = (headroom > 16 ? headroom : 16) - skb_headroom(skb);
  949. if (delta < 0)
  950. delta = 0;
  951. if (delta || skb_cloned(skb))
  952. return pskb_expand_head(skb, (delta + 15) & ~15, 0, GFP_ATOMIC);
  953. return 0;
  954. }
  955. /**
  956. * skb_padto - pad an skbuff up to a minimal size
  957. * @skb: buffer to pad
  958. * @len: minimal length
  959. *
  960. * Pads up a buffer to ensure the trailing bytes exist and are
  961. * blanked. If the buffer already contains sufficient data it
  962. * is untouched. Returns the buffer, which may be a replacement
  963. * for the original, or NULL for out of memory - in which case
  964. * the original buffer is still freed.
  965. */
  966. static inline struct sk_buff *skb_padto(struct sk_buff *skb, unsigned int len)
  967. {
  968. unsigned int size = skb->len;
  969. if (likely(size >= len))
  970. return skb;
  971. return skb_pad(skb, len-size);
  972. }
  973. static inline int skb_add_data(struct sk_buff *skb,
  974. char __user *from, int copy)
  975. {
  976. const int off = skb->len;
  977. if (skb->ip_summed == CHECKSUM_NONE) {
  978. int err = 0;
  979. unsigned int csum = csum_and_copy_from_user(from,
  980. skb_put(skb, copy),
  981. copy, 0, &err);
  982. if (!err) {
  983. skb->csum = csum_block_add(skb->csum, csum, off);
  984. return 0;
  985. }
  986. } else if (!copy_from_user(skb_put(skb, copy), from, copy))
  987. return 0;
  988. __skb_trim(skb, off);
  989. return -EFAULT;
  990. }
  991. static inline int skb_can_coalesce(struct sk_buff *skb, int i,
  992. struct page *page, int off)
  993. {
  994. if (i) {
  995. struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
  996. return page == frag->page &&
  997. off == frag->page_offset + frag->size;
  998. }
  999. return 0;
  1000. }
  1001. /**
  1002. * skb_linearize - convert paged skb to linear one
  1003. * @skb: buffer to linarize
  1004. * @gfp: allocation mode
  1005. *
  1006. * If there is no free memory -ENOMEM is returned, otherwise zero
  1007. * is returned and the old skb data released.
  1008. */
  1009. extern int __skb_linearize(struct sk_buff *skb, unsigned int __nocast gfp);
  1010. static inline int skb_linearize(struct sk_buff *skb, unsigned int __nocast gfp)
  1011. {
  1012. return __skb_linearize(skb, gfp);
  1013. }
  1014. /**
  1015. * skb_postpull_rcsum - update checksum for received skb after pull
  1016. * @skb: buffer to update
  1017. * @start: start of data before pull
  1018. * @len: length of data pulled
  1019. *
  1020. * After doing a pull on a received packet, you need to call this to
  1021. * update the CHECKSUM_HW checksum, or set ip_summed to CHECKSUM_NONE
  1022. * so that it can be recomputed from scratch.
  1023. */
  1024. static inline void skb_postpull_rcsum(struct sk_buff *skb,
  1025. const void *start, int len)
  1026. {
  1027. if (skb->ip_summed == CHECKSUM_HW)
  1028. skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
  1029. }
  1030. /**
  1031. * pskb_trim_rcsum - trim received skb and update checksum
  1032. * @skb: buffer to trim
  1033. * @len: new length
  1034. *
  1035. * This is exactly the same as pskb_trim except that it ensures the
  1036. * checksum of received packets are still valid after the operation.
  1037. */
  1038. static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
  1039. {
  1040. if (len >= skb->len)
  1041. return 0;
  1042. if (skb->ip_summed == CHECKSUM_HW)
  1043. skb->ip_summed = CHECKSUM_NONE;
  1044. return __pskb_trim(skb, len);
  1045. }
  1046. static inline void *kmap_skb_frag(const skb_frag_t *frag)
  1047. {
  1048. #ifdef CONFIG_HIGHMEM
  1049. BUG_ON(in_irq());
  1050. local_bh_disable();
  1051. #endif
  1052. return kmap_atomic(frag->page, KM_SKB_DATA_SOFTIRQ);
  1053. }
  1054. static inline void kunmap_skb_frag(void *vaddr)
  1055. {
  1056. kunmap_atomic(vaddr, KM_SKB_DATA_SOFTIRQ);
  1057. #ifdef CONFIG_HIGHMEM
  1058. local_bh_enable();
  1059. #endif
  1060. }
  1061. #define skb_queue_walk(queue, skb) \
  1062. for (skb = (queue)->next; \
  1063. prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
  1064. skb = skb->next)
  1065. extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
  1066. int noblock, int *err);
  1067. extern unsigned int datagram_poll(struct file *file, struct socket *sock,
  1068. struct poll_table_struct *wait);
  1069. extern int skb_copy_datagram_iovec(const struct sk_buff *from,
  1070. int offset, struct iovec *to,
  1071. int size);
  1072. extern int skb_copy_and_csum_datagram_iovec(const
  1073. struct sk_buff *skb,
  1074. int hlen,
  1075. struct iovec *iov);
  1076. extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
  1077. extern unsigned int skb_checksum(const struct sk_buff *skb, int offset,
  1078. int len, unsigned int csum);
  1079. extern int skb_copy_bits(const struct sk_buff *skb, int offset,
  1080. void *to, int len);
  1081. extern int skb_store_bits(const struct sk_buff *skb, int offset,
  1082. void *from, int len);
  1083. extern unsigned int skb_copy_and_csum_bits(const struct sk_buff *skb,
  1084. int offset, u8 *to, int len,
  1085. unsigned int csum);
  1086. extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
  1087. extern void skb_split(struct sk_buff *skb,
  1088. struct sk_buff *skb1, const u32 len);
  1089. static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
  1090. int len, void *buffer)
  1091. {
  1092. int hlen = skb_headlen(skb);
  1093. if (hlen - offset >= len)
  1094. return skb->data + offset;
  1095. if (skb_copy_bits(skb, offset, buffer, len) < 0)
  1096. return NULL;
  1097. return buffer;
  1098. }
  1099. extern void skb_init(void);
  1100. extern void skb_add_mtu(int mtu);
  1101. #ifdef CONFIG_NETFILTER
  1102. static inline void nf_conntrack_put(struct nf_conntrack *nfct)
  1103. {
  1104. if (nfct && atomic_dec_and_test(&nfct->use))
  1105. nfct->destroy(nfct);
  1106. }
  1107. static inline void nf_conntrack_get(struct nf_conntrack *nfct)
  1108. {
  1109. if (nfct)
  1110. atomic_inc(&nfct->use);
  1111. }
  1112. static inline void nf_reset(struct sk_buff *skb)
  1113. {
  1114. nf_conntrack_put(skb->nfct);
  1115. skb->nfct = NULL;
  1116. }
  1117. #ifdef CONFIG_BRIDGE_NETFILTER
  1118. static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
  1119. {
  1120. if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
  1121. kfree(nf_bridge);
  1122. }
  1123. static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
  1124. {
  1125. if (nf_bridge)
  1126. atomic_inc(&nf_bridge->use);
  1127. }
  1128. #endif /* CONFIG_BRIDGE_NETFILTER */
  1129. #else /* CONFIG_NETFILTER */
  1130. static inline void nf_reset(struct sk_buff *skb) {}
  1131. #endif /* CONFIG_NETFILTER */
  1132. #endif /* __KERNEL__ */
  1133. #endif /* _LINUX_SKBUFF_H */